JP2005298223A - Silicon single crystal growing method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method by which a high quality silicon single crystal having a large diameter in which grown-in defects are reduced as much as possible, can be manufactured in a good yield at high pulling/growing speed. <P>SOLUTION: (1) In the production of the silicon single crystal for a semiconductor by a CZ method, the method for growing the silicon single crystal is characterized by pulling the single crystal while applying a horizontal magnetic field to a melt in a crucible so that the center of the magnetic field (the position of the maximum value of the magnetic flux density) is situated within the range of 100-600 mm in the depth from the surface of the melt, and the fluctuation of the magnetic flux density distribution is ≥3% in the range. (2) Further, the method for growing the silicon single crystal mentioned in (1) is characterized in that the magnetic field intensity at the center of the magnetic field is within the range of 0.1-0.2 T. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a method for growing a silicon single crystal used as a semiconductor material by the Czochralski method (hereinafter referred to as “CZ method”). More specifically, a high-quality crystal with few crystal defects can be efficiently produced. The present invention relates to a silicon single crystal growth method.

  The most widely adopted method for producing a silicon single crystal used for a semiconductor material silicon wafer is a CZ method.

  In the CZ method, a single crystal is grown by immersing a seed crystal in molten silicon in a quartz crucible and growing the single crystal. Due to the advancement of this silicon single crystal growth technology, a dislocation-free single crystal with few defects is produced. It has come to be. A semiconductor device is manufactured as a substrate after several hundred processes using a wafer obtained from a single crystal as a substrate.

  In the process, the substrate is subjected to many physical treatments, chemical treatments, and thermal treatments, and includes processing in a severe thermal environment such as high-temperature treatment at 1000 ° C. or higher. For this reason, the cause is introduced in the growth process of the single crystal, and the micro defect, that is, the Grown-in defect, which becomes obvious in the manufacturing process of the device and results in the deterioration of the performance becomes a problem.

  FIG. 1 is a diagram schematically showing an example of a typical defect distribution observed on a silicon wafer. The distribution of representative microdefects shown in the figure is that a wafer is cut out from a single crystal immediately after growth, dipped in an aqueous copper nitrate solution to deposit Cu, and after heat treatment, the microdefect distribution is observed by X-ray topography. The results are shown schematically.

  That is, in this wafer, an oxidation-induced stacking fault distributed in a ring shape (hereinafter referred to as OSF (Oxygen Induced Stacking Fault)) appears at a position about 2/3 of the outer diameter. Infrared scatterer defects (also referred to as COP or FPD, both of which are defective in the same Si state) are found. In addition, there is an oxygen precipitation promoting region where oxygen precipitates are likely to appear immediately outside the ring-shaped OSF, and there is a defect-free region where defects do not appear, and dislocation clusters are generated around the outer periphery of the wafer. It is easy to do.

  The position where the above defects are generated is usually greatly affected by the pulling speed during single crystal growth. As an example, a single crystal grown while continuously reducing the pulling rate within the range of the growth rate to obtain a healthy single crystal without dislocation was cut longitudinally along the pulling axis at the center of the crystal. When the distribution of various defects on the surface is examined, a result as shown in FIG. 2A is obtained.

  When viewed on a disk-shaped wafer surface cut perpendicular to the single crystal pulling axis, a ring-shaped OSF appears from the periphery of the crystal if the growth rate is lowered after forming the shoulder portion to obtain the required straight body diameter. . The ring-shaped OSF that appears in the peripheral portion gradually decreases in diameter as the growth rate decreases, and eventually disappears, and the entire surface of the wafer corresponds to the outer portion of the ring-shaped OSF.

  That is, FIG. 1 shows a cross section perpendicular to the single crystal A pulling axis in FIG. 2A, or a single crystal wafer grown at the pulling speed, and shows the position of occurrence of the ring-shaped OSF. As a standard, when the growth rate is fast, the single crystal grows relatively fast with a large number of infrared scatterer defects corresponding to the inner region of the ring-shaped OSF, and when slow, the slow growth single crystal has many dislocation clusters in the outer region. It becomes.

  When the melted silicon solidifies into a single crystal, the defect part lacking silicon atoms as the crystal lattice and the excess defect part are taken in at the same time, and these coalesce or diffuse and disappear, Infrared scatterer defects are left in the finally missing part, and dislocation clusters remain in the excess part. Then, a portion where both are not excessive or deficient becomes a defect-free region, and it is considered that a ring-shaped OSF appears at a specific position in the region, and this defect-free region is sometimes called a neutral region.

  It is well known that dislocations in a silicon single crystal cause deterioration of the characteristics of a device formed thereon. In addition, the OSF deteriorates electrical characteristics such as an increase in leakage current, but the ring-shaped OSF has a high density.

  Therefore, at present, for a normal LSI, a single crystal is grown at a relatively high pulling speed such that the ring-shaped OSF is distributed on the outermost periphery of the single crystal. Thereby, most of the wafer is made into an inner part of the ring-shaped OSF, that is, a high-speed grown single crystal, thereby avoiding dislocation clusters. This is because the inner part of the ring-shaped OSF has a larger gettering action against heavy metal contamination generated in the manufacturing process of the device than the outer part.

  In recent years, with the increase in the degree of integration of LSI, the gate oxide film has been made thinner, and the temperature in the device manufacturing process has been lowered. For this reason, the OSF that is likely to be generated by high-temperature treatment is reduced, and the crystal has low oxygen, so that the OSF such as a ring-shaped OSF has less problems as a factor that deteriorates the device characteristics.

  However, it has been clarified that the presence of infrared scatterer defects mainly present in high-speed grown single crystals greatly deteriorates the breakdown voltage characteristics of the thinned gate oxide film, and in particular, the device pattern becomes finer. As a result, the effect becomes large and it becomes difficult to cope with high integration.

  In the defect distribution shown in FIG. 1, if the oxygen precipitation promoting region and the defect-free region in contact with the ring-shaped OSF can be enlarged, a wafer or a single crystal with very few grown-in defects may be obtained.

  For example, Patent Document 1 discloses that the pulling speed during single crystal growth is V (mm / min), and the temperature gradient in the pulling axis direction in the temperature range from the melting point to 1300 ° C. is G (° C./mm). In the internal position from the outer periphery to 30 mm from the center of the crystal, V / G is set to 0.20 to 0.22, and the temperature gradient in the crystal is controlled so as to gradually increase toward the outer periphery of the crystal, thereby dislocation clusters. There has been proposed a method in which only the defect-free region of the outer portion of the ring-shaped OSF is extended to the entire wafer surface or even the entire single crystal without generating.

  As described above, when a single crystal is grown while the pulling rate is continuously decreased, the ring-shaped OSF usually appears in a V shape shown in FIG. On the other hand, the method disclosed in Patent Document 1 is such that the temperature gradient in the pulling-up axis direction is such that the central part of the crystal is large and the peripheral part is small, and the range is limited, so that FIG. The generation of the single crystal ring-shaped OSF grown in the same manner as described above can be changed to a U-shape or a flat bottom as shown in FIG.

Therefore, for example, if the growth is performed at the pulling rate shown as B ₁ , a single crystal in which the defect-free region is greatly expanded can be obtained. In the method disclosed in Patent Document 1, the temperature gradient G in the crystal is obtained and predicted by heat transfer analysis simulation, but the specific method for realizing such a temperature gradient is not necessarily clear.

  Thereafter, several manufacturing methods have been proposed for realizing the idea of making the entire single crystal a defect-free region. For example, in Patent Document 2, a heat insulating material is disposed so as to surround a silicon single crystal immediately above the molten metal surface of the silicon melt during single crystal growth, and the gap between the heat insulating material and the molten metal surface is set to 30 to 50 mm. Using a device that applies a horizontal magnetic field of 2 T or more, the shape of the solid-liquid interface of the crystal is within ± 5 mm of the average value, and the temperature gradient in the pulling axis direction is 0. An invention of a manufacturing method has been disclosed in which the entire single crystal is brought into the above-described defect-free region or neutral state region by adjusting the pulling speed within 5 ° C./mm.

  Next, in Patent Document 3, as a method for manufacturing the defect-free silicon single crystal, it is preferable to make the shape of the solid-liquid interface flat or upwardly convex. There has been proposed an invention in which the speed is increased at 5 rotations / minute or less and the single crystal is pulled at a rotation speed of 13 rotations / minute or more.

  Further, in Patent Document 4, by selecting a rotation speed of the crucible and the single crystal, setting a position of the heat shield disposed around the single crystal, and applying a horizontal magnetic field or a cusp magnetic field to the melt, etc., The solid-liquid interface shape is an upward convex shape with a central part height exceeding 5 mm with respect to the peripheral part, and the temperature gradient in the pulling axis direction of the single crystal is set as a temperature distribution in which the central part is large and the peripheral part is small. An invention of a method for producing a high-quality silicon single crystal that is entirely defect-free is disclosed.

  Further, Patent Document 5 proposes an invention for obtaining the above defect-free single crystal by controlling the temperature distribution on the side surface of the single crystal as well as the shape of the upward convex solid-liquid interface.

  With the development of multi-function control of electronic equipment, the LSI chip size has been increased, but in order to increase the chip collection rate per wafer, the diameter of the wafer, that is, the diameter of the silicon single crystal has been increased. Necessary. Further, increasing the chip size reduces the yield of non-defective products unless the density of point defects such as grown-in defects is made lower than before. For this reason, the single crystal which expanded the defect-free area | region is desirable.

  However, when the diameter of the single crystal is increased to 300 mm or more, it is not easy to remove heat from the single crystal, and if an attempt is made to enhance cooling to improve productivity, the thermal stress increases and dislocations are generated. Or cracks may occur in the crystal. For this reason, the pulling speed cannot be increased, and the productivity is greatly reduced.

  In addition to this, when trying to obtain a silicon single crystal with an expanded defect-free region, the temperature gradient in the single crystal is controlled as in the conventional single crystal of 200 mm or smaller when it becomes a large single crystal. It is not easy to do so, and the pulling speed range in which a defect-free single crystal can be grown becomes extremely narrow, making it difficult to control the speed and greatly reducing the yield of defect-free portions.

JP-A-8-330316

JP 11-79889 A JP 2000-72590 A JP 2001-158690 A JP 2001-261495 A JP 2000-7485 A

  An object of the present invention is to provide a method capable of producing a high-quality crystal having as few as possible grown-in defects in a large-diameter silicon single crystal having a diameter of 300 mm or more with a high yield and a high growth rate.

  As described above, conventionally, the temperature distribution in the single crystal and the shape of the solid-liquid interface at the time of pulling are controlled by using a heat insulating material, controlling the rotation of the crucible and the single crystal, and applying a magnetic field to the melt. There has been proposed a method for producing a single crystal in which the pulling speed is selected to make the whole defect-free region.

  However, as the diameter of the single crystal increases, the pulling speed must be greatly reduced, and the optimum speed range for defect-free is narrowed, making it difficult to apply to actual production. It has become.

  On the other hand, the present invention provides a method for producing a defect-free single crystal capable of expanding the optimum speed range and increasing the growth speed.

  In the present invention, (1) in the production of a silicon single crystal for a semiconductor that pulls up a single crystal while applying a horizontal magnetic field to the melt in the crucible by the CZ method, the depth from the melt surface is 100 mm to 600 mm. The silicon single crystal is characterized in that the magnetic field center (the maximum value position of the magnetic flux density) is located in the range and the pulling is performed while applying a magnetic field in which the fluctuation of the magnetic flux density distribution in the range is 3% or more. (1) The method for growing a silicon single crystal according to (1) above, wherein the magnetic field strength at the center of the magnetic field is in the range of 0.1 to 0.2T.

  In a large-diameter silicon single crystal, a high-quality crystal with as few grown-in defects as possible can be manufactured with a high yield and a high pulling growth rate. A wafer obtained from such a single crystal can be effectively applied to increase the yield and reduce the cost in response to the trend toward higher density and larger LSI chips.

In order to produce a single crystal having as few as possible a grown-in defect such as a dislocation cluster defect or an infrared scatterer defect, or a point defect, a ring-like OSF or non-destructive crystal can be obtained when grown while the pulling rate is continuously reduced. The defect area is made flat as shown in FIG. 2B, and then B ₂ (infrared scatterer) from B velocity or B ₁ (lower pulling rate at which dislocation cluster defects do not appear). It is grown at a speed between the upper limit pulling speed at which no defect appears).

  In order to make the defect-free region having the shape as shown in FIG. 2B appear, in the conventional single crystal having an outer diameter of up to 200 mm, for example, as shown in the aforementioned Patent Document 4, the single crystal being grown In this method, the temperature gradient in the crystal in the pulling axis direction is made large at the center and small at the periphery, and the solid-liquid interface shape is made as an upward convex shape, and the pulling rate is limited and growing.

  However, when the diameter of the single crystal is 300 mm or more, it is not easy to find a pulling speed at which the entire wafer surface becomes a defect-free region. Even if it can be realized, the allowable speed range is extremely small and stable and defect-free. It becomes difficult to obtain a single crystal. The defect-free region shown in FIG. 2B is not sufficiently flat in the horizontal direction.

  This is because the larger the diameter of the single crystal, the more difficult it is to remove heat from the crystal immediately after solidification, and the use of a heat insulating material or a cooling body, or gas blowing, etc. This is because the temperature distribution control is not sufficient. If this control is performed more strongly, the thermal strain due to the temperature difference increases, and there is a risk of causing dislocations and further crystal cracks.

  On the other hand, in order to increase the height of the solid-liquid interface shape with an upward convex shape, there is a method of increasing the rotational speed of the single crystal and further applying a magnetic field to the melt to control the convection of the melt. Increasing the rotation speed of the single crystal cannot be sufficiently increased because dislocations are likely to occur when the diameter increases.

  Thus, when the diameter of the single crystal is increased, the temperature distribution in the pulling axis direction inside the single crystal is not realized sufficiently, and the temperature distribution corresponding to the diameter of the single crystal that the central portion is large and the peripheral portion is small, As shown in FIG. 2B, a flat bottom-like defect-free region cannot be obtained sufficiently.

  However, since the temperature of the solidification interface is constant, if the upward convex shape of the solid-liquid interface can be made larger, the temperature gradient in the pulling axial direction at the center will be that of the periphery when the temperature distribution in the single crystal being pulled is similar. It is thought that there is a possibility that it can be made larger.

  The shape of the solid-liquid interface has a constant latent heat of solidification, so it is determined by the balance between the supply of heat from the melt and the removal of heat by heat transfer to the solid crystal. Therefore, the shape of the solid-liquid interface should be controlled by controlling the supply of heat from the melt, that is, the convection of the melt.

  It is the rotation of the crystal and the application of the magnetic field that greatly affect the flow of the melt, such as convection.However, as the diameter of the single crystal increases, the rotation of the crystal is limited as described above. I tried various investigations.

  As a result, it was found that the shape of the solid-liquid interface, that is, the height of the upwardly convex shape can be effectively increased by applying a specific magnetic field in a range limited to the melt. Therefore, the relationship between the shape of the solid-liquid interface, the generation shape of the defect-free region of the single crystal obtained by continuously changing the pulling rate, and the conditions under which a healthy single crystal without dislocations can be grown is further It was confirmed that the solid-liquid interface shape has a preferable range.

  Conventionally, magnetic field application during silicon single crystal growth is mainly used to control the amount of oxygen taken into the single crystal by suppressing convection of the melt, and may affect the solid-liquid interface shape during single crystal growth. Are known. In the horizontal magnetic field application that is normally used, a magnetic field with good uniformity is applied to the melt. In the horizontal magnetic field, the magnetic flux density at the central axis position of the magnetic field generating coil is usually the highest, and the magnetic flux density decreases as the distance from the center increases. To do.

  Therefore, the rotation speed of the single crystal was set near the upper limit for stable growth, and the influence of the strength of the magnetic field and the position of the magnetic field center on the solid-liquid interface shape was examined. As a result, if the strength of the magnetic field is increased, the upward convex shape is suppressed, and if the magnetic field strength is decreased, the height of the upward convex shape can be increased. It was found that the center of the magnetic field should be at least 100 mm deep from the melt surface.

  In the course of such an investigation, when the magnetic field generator used for applying the magnetic field was replaced and tested, it was found that using a non-uniform magnetic field was more effective than using a uniform magnetic field. .

  Therefore, the limit of these conditions, with the goal that the temperature gradient in the pulling axis direction in the single crystal is large in the central part and small in the peripheral part, and the defect-free region in FIG. Further study was conducted.

  As a result, the applied magnetic field is a horizontal magnetic field, the magnetic field center (maximum magnetic flux density position) is located in the range from 100 mm to 600 mm in depth from the melt surface, and the fluctuation of the magnetic flux density distribution in that range. It has been confirmed that it is good to be 3% or more. More desirably, the magnetic field strength at the center of the magnetic field is in the range of 0.1 to 0.2T.

  Various attempts were made to grow single crystals under such conditions, and it was found that the pulling speed could be increased and the allowable speed range (speed margin) for producing defect-free single crystals could be greatly expanded.

  So far, when the diameter is 300 mm, when trying to obtain a single crystal that is defect-free with respect to a grown-in defect, the speed margin is around 0.01 mm, and it is extremely difficult to expand the defect-free region throughout the single crystal. It was. However, if the growth is performed under the above conditions, this speed margin is nearly doubled. In addition, the pulling speed for making a defect-free region can be improved by 10 to 40%.

  Here, when the position of the magnetic field center is shallower than 100 mm from the melt surface, the upwardly convex shape of the solid-liquid interface becomes insufficient. This seems to be because the flow of the melt is suppressed. Also, when the position of the magnetic field center is deeper than 600 mm from the melt surface, dislocations and cracks of the single crystal may occur, but this is because the convex shape of the solid-liquid interface becomes too high. Seem.

  When the magnetic field strength at the magnetic field center is smaller than 0.1T, the upwardly convex shape becomes too high. Moreover, if it is less than 0.2T, the height of the upward convex shape will be low. For this reason, it is desirable that the magnetic field strength at the center of the magnetic field be in the range of 0.1 to 0.2T.

  Further, when the magnetic flux density of the magnetic field is non-uniform rather than uniform, the defect-free region of the single crystal is expanded. The reason for this is not necessarily clear, but it is considered that the non-uniformity facilitates the flow of the melt and makes the upwardly convex shape more favorable.

When this non-uniform state is defined more clearly, if the maximum magnetic flux density T _H and the minimum magnetic flux density T _L are within the range from 100 mm to 600 mm in depth from the melt surface, the following equation (1) ΔT (%) indicating the non-uniformity indicated by is a value exceeding 3%.

ΔT (%) = 100 × (T _H −T _L ) / T _H (1)
From the viewpoint of suppressing the occurrence of dislocations and cracks in the single crystal, ΔT is preferably kept within 10% at most, and the magnitude of ΔT can be changed by changing the vertical coil interval of the horizontal magnetic field generating coil. Can be adjusted.

  The application of the magnetic field as described above can be realized by arranging the magnetic field central axis of the horizontal magnetic field generating coil having a non-uniform magnetic field in the range from 100 mm to 600 mm in depth from the melt surface. In this case, it is more preferable to place the magnetic field center at a depth in the range of 200 mm to 400 mm. This is because the effect of making the solid-liquid interface convex upward is greater when the depth exceeds 200 mm, and the instability increases without much change beyond 400 mm.

  The apparatus used for applying the magnetic field is not particularly limited as long as it can generate the magnetic field. However, as the diameter of the single crystal increases, the magnetic field generator increases in size. The increase in size of the magnetic field generator makes it difficult to attach and detach the work around the furnace for melting and pulling up, such as replacing the crucible of the single crystal manufacturing apparatus.

  On the other hand, it is assumed that the magnetic coil that generates the magnetic field can be easily divided into two as a saddle type, and both the vertical magnetic field and the horizontal magnetic field can be generated by changing the distance between the two coils. It is disclosed in Patent Document 6. If such a saddle-shaped magnetic coil is used to generate a horizontal magnetic field, the above-mentioned non-uniform magnetic field can be easily obtained.

  The rotation of the single crystal and the rotation of the crucible during the growth are not particularly limited, but the convection of the melt is suppressed by a magnetic field, while the flow of the melt is larger with respect to the solid-liquid interface of the single crystal. It is preferable that the surface is smaller than the crucible surface. Therefore, the rotation of the single crystal is preferably 6 rotations / minute or more and the crucible rotation is 5 rotations / minute or less.

  Using a single crystal growth apparatus having a diameter of 300 mm, 230 kg of polycrystalline silicon was added in a quartz crucible having an inner diameter of 810 mm so as to have an electric resistance of 20 Ωcm, and was melted in an argon atmosphere at 2600 Pa. The application of the magnetic field was a horizontal magnetic field, and the generator used was one that generates a conventional uniform magnetic field or one that can generate a non-uniform magnetic field. In the case of a uniform magnetic field, ΔT in the above formula (1) is 2.3%. In the case of a non-uniform magnetic field, the magnetism generator is of a saddle type coil, and ΔT is 3%, 6% and 7%. did.

  The seed crystal was brought into contact with the melt and pulled up to form predetermined neck portions and shoulder portions. After reaching the target body diameter, the pulling rate was gradually reduced to grow a silicon single crystal. The rotational speed of the single crystal was 13 times / minute, and the rotational speed of the crucible was 3 times / minute. In order to confirm the shape of the solid-liquid interface, after raising the pulling speed expected to form a defect-free region, the growth was stopped and the single crystal was pulled away from the melt.

  About the bottom face of the obtained single crystal, the height of the center part with respect to the peripheral part was measured, and the rate of increase of the solid-liquid interface was measured. In this case, since the diameter of the single crystal varies slightly, the ratio of the height of the central portion to the single crystal diameter is obtained, and when the uniform magnetic field is applied and the magnetic field center is positioned 50 mm below the melt surface The ratio was set to 100 (%), and in the case of other conditions, the ratio to this was set as the rate of increase.

  Further, the single crystal was cut in parallel with the pulling shaft, and a slice piece having a thickness of 1.4 mm including the pulling shaft was collected and immersed in a 16% by mass copper nitrate solution to attach copper, at 900 ° C. After heating for 20 minutes and cooling, the defect distribution was investigated by X-ray topography.

  Table 1 shows the increase in the solid-liquid interface height when a normal horizontal magnetic field and an inhomogeneous magnetic field with ΔT of 3%, 6%, and 7% are applied at different heights from the melt surface at the center of the magnetic field. Indicates the rate. It can be seen that the solid-liquid interface height is significantly increased in the non-uniform magnetic field as compared with the case of the normal horizontal magnetic field. However, when the solid-liquid interface height greatly exceeds 50 mm, dislocation is likely to occur, which is not preferable.

  Table 2 shows the growth rate improvement rate and the growth rate margin determined from the distribution of defect-free regions when the magnetic field center position where the magnetic flux density at the magnetic field center is 0.2T and the non-uniformity ΔT is 6% is changed. (Speed range for obtaining a defect-free single crystal) is shown.

  The growth speed is based on the speed when a defect-free single crystal is produced by applying a normal uniform horizontal magnetic field, but the growth speed is improved if the center position of the magnetic field is −50 mm even if a non-uniform magnetic field is used. This is almost the same as applying a uniform magnetic field. It can be seen that the growth rate margin was about 0.01 mm when a normal horizontal magnetic field was applied, but the range was expanded if a non-uniform magnetic field was applied.

  Table 3 shows the results when the position of the magnetic field center is −280 mm from the melt surface and the intensity of the inhomogeneous magnetic field of ΔT = 6% is changed. When the strength of the magnetic field is less than 0.1T, the rise in the solid-liquid interface height is slightly too high, and when it exceeds 0.2T, the solid-liquid interface height does not rise sufficiently.

It is the figure which showed typically the example of the typical defect distribution observed with a silicon wafer. It is the figure which showed typically distribution of the defect in the longitudinal cross-section of the single crystal grown by reducing the pulling rate continuously. {(A) Normal case corresponding to FIG. 1 (b) Controlling temperature distribution in a single crystal}

Claims (2)

In the production of a silicon single crystal for semiconductors that pulls up the single crystal while applying a horizontal magnetic field to the melt in the crucible by the Czochralski method,
A magnetic field in which the center of the magnetic field (maximum magnetic flux density position) is located in a range from 100 mm to 600 mm in depth from the melt surface, and a magnetic flux density distribution variation in the range is 3% or more is applied. A method for growing a silicon single crystal, wherein the silicon single crystal is pulled up while being pulled. 2. The method for growing a silicon single crystal according to claim 1, wherein the magnetic field strength at the magnetic field center is in the range of 0.1 to 0.2T.

Images